[LMCE 2026] Comparative Analytical Evaluation of Automated Pre-classification by Two Digital Morphology Analyzers against Manual Microscopy for Leukocyte Differential and Abnormal Cell Detection

miLab™ BCM showed differential performance comparable to DC-1, with far fewer false-positive IG flags that may reduce unnecessary manual review. Further validation is warranted.

LMCE 2026 (Laboratory Medicine Congress & Exhibition, Seoul)

October 21, 2026

C.Y. Bae, et al.

[LMCE 2026] Comparative Analytical Evaluation of Automated Pre-classification by Two Digital Morphology Analyzers against Manual Microscopy for Leukocyte Differential and Abnormal Cell Detection

Unauthorized downloading or distribution of this poster is strictly prohibited.

Abstract

This study compared the stand-alone AI pre-classification of miLab™ BCM and CellaVision DC-1 against manual microscopy in 292 paired samples. miLab™ BCM showed closer lymphocyte agreement and far fewer false-positive IG flags, with comparable blast discrimination.

AUC = area under the ROC curve; NPA = negative percent agreement at 100% positive percent agreement; FP = false-positive flags among reference-negative samples. 292 paired samples.

Key Highlights

Study at Asan Medical Center (Seoul): 302 samples, 293 analyzed (292 paired), with the mean of two independent 200-cell manual counts as reference.
Both analyzers were assessed on stand-alone AI pre-classification (200 cells, before reviewer verification); miLab™ BCM analyzed slides prepared on its own platform.
Lymphocyte: 99.0% of miLab™ BCM results fell within the acceptable limit vs 81.5% for DC-1 (Wilcoxon P<0.001).
IG ≥4% flag: AUC 0.997 vs 0.762; NPA at 100% PPA 97.9% vs 30.3% (6 vs 198 false positives).
Blast ≥1% flag: similar AUC (0.987 vs 0.972); NPA was lower for miLab™ BCM (80.7% vs 88.2%).
Specimen collection, automated slide preparation and manual review were performed independently at Asan Medical Center.

Download Brochure Request Demo